1. Field of the Invention
The present invention generally relates to charged-particle-beam exposure devices, and particularly relates to a charged-particle-beam exposure device which forms a pattern on a wafer by exposing the wafer to charged particles.
2. Description of the Related Art
As the circuit density of semiconductor integrated circuits increases, a finer processing technique is required. Compared to the light exposure method widely used in the manufacturing of LSI chips, the charged-particle exposure method has much superior characteristics in terms of the resolution and the focus depth. With respect to the resolution, a processing limit of the photolithography method is about 0.3 .mu.m, while processing as fine as 0.1 .mu.m can be achieved in the charged-particle-beam exposure method.
However, the charged-particle-beam exposure method is inferior compared to the light exposure method in terms of an exposure positioning accuracy, an overlay accuracy, and a field stitching accuracy. Because of this, the charged-particle-beam exposure method is not widely used in the field for manufacturing purposes of LSI chips.
The charged-particle-beam exposure device has a smaller area to be able to be exposed at one time, compared to a light exposure device such as a stepper. (This area is called a deflection field hereinafter.) Thus, in order to expose one LSI chip, stage movement is required to successively shift the deflection field on the LSI chip. In doing so, if the connecting precision across borders of the deflection fields is low, severance of wires and/or short-circuits are generated which greatly degrades the yield of the chips.
In order to improve the yield, the connecting precision at the field borders must be enhanced, which requires a higher precision of the deflection of the charged-particle beam. In the charged-particle-beam exposure device, the charged-particle beam is generally deflected by a magnetic field generated by coils. The coils include two systems for x-direction deflection and for y-direction deflection. Separate currents are applied to these two systems to deflect the beam in the x direction and the y direction independently. Unfortunately, the amount of beam deflection is not in proportion to the amount of current applied to the deflection coils, but is represented as a complex function of the current amount.
In order to deflect the beam with a high precision, the amount of the current applied to the deflector must be corrected. There are two types of corrections. One is a distortion correction for establishing a linear relation between the input and the deflection amount, and the other is a deflection-efficiency correction for correcting coefficients for linear factors. The distortion correction is a time consuming process since it requires data collection at various points within the field. However, the data needs to be collected only one time since a time variation of the distortion is small. On the other hand, correction coefficients can be obtained in a short period of time for the deflection-efficiency correction. However, the deflection-efficiency-correction coefficients must be frequently obtained because the deflection efficiency varies over time due to a change in thermal distribution of the deflector, etc.
In order to calibrate the deflection field, coordinates of the deflector are generally matched with coordinates of the stage, whose measure and orthogonality are guaranteed through the laser-interferometer system. In order to measure the coordinates of the deflector, an actual position of the charged-particle beam must be obtained by directing the beam to mark positions on a wafer and detecting reflected charged particles.
FIG. 1A is an illustrative drawing for explaining a method of detecting mark positions through the charged-particle-beam scan. As shown in FIG. 1A, the charged-particle beam is scanned by the deflector over a mark 306 formed as a groove in a reference chip 305. Reflection detectors 300 and 301, symmetrically arranged with respect to the axis of the beam optical system, detect reflected charged particles. Outputs of the detectors are added by the adder 302. A signal after the addition is successively obtained in synchronism with the scan of the deflector, providing a reflection signal form to be analyzed. When such a process is conducted by using the position-detection mark 306 as shown in FIG. 1A, a reflection signal form as shown in FIG. 1B is obtained. The reflection signal form obtained in this manner is analyzed by an analyzing device 303 to detect a center position of the mark. A result of the analysis is sent from the analyzing device 303 to a control-purpose computer 304, which uses the result in processes such as a correction of the beam. In general, a groove (dent) formed in a wafer (silicon) is used as a mark.
The detection of the position mark described above is conducted at various points by shifting the mark on a wafer through stage movement. In this manner, the deflection-efficiency-correction coefficients for correcting the linear factors and a distortion map of the deflector for the distortion correction are obtained.
In the mark-position-detection method described above, the detected mark positions contain errors. This is because a relative position of the mark with respect to the reflection detectors changes when the mark is detected at various points.
When the mark is detected at various points, an angle at which charged particles are reflected by the mark toward a reflection detector varies depending on a relative position of the mark with respect to the reflection detector. When reflected charged particles are detected in a configuration as shown in FIG. 2A, signal forms as shown in FIG. 2B are obtained. As shown in figures, a reflection signal having a symmetric form without a distortion can be obtained when the mark is positioned at an equal distance from the two reflection detectors. When the mark is positioned at other locations, however, a reflection signal form having an asymmetry is obtained. This is because the angle of the reflection is different for the different reflection detectors.
In addition to the problems of errors regarding the mark-position detection, there is a problem concerning the focusing of the charged-particle beam in the charged-particle-beam exposure device.
FIG. 3 is an illustrative drawing showing a configuration for the focusing of the beam in the related-art charged-particle-beam exposure device. As shown in FIG. 3, an optical system 310, using a type of light not affecting a resist, is provided between a wafer and a charged-particle lens. The optical system 310 includes a light source 311 and a light detector 312. When the wafer is exposed to the charged-particle beam, the light source 311 illuminates light on the wafer, and the light detector 312 detects light reflected from the wafer to measure the height of an exposed surface. Based on the height of the exposed surface, a focusing distance of the charged-particle lens is changed.
Such a related-art charged-particle-beam exposure device has such problems as:
a) when the focusing distance of the reflection path is changed, the deflection path of the charged-particle beam is affected to cause a displacement of the beam position on the wafer surface; and PA1 b) since structures under the exposed surface have complex patterns in a LSI device, light reflected from these patterns has an adverse effect of causing errors in the detection of the height.
The problem a) will be described below. In the charged-particle-beam exposure device, deflection coordinates X=(X, Y), having an origin at the axis of the beam optical system, are entered into a correction circuit to obtain corrected deflection coordinates X'=(X', Y'). EQU X'=Gx.cndot.X+Rx.cndot.Y+Dx(X, Y) (1) EQU Y'=Ry.cndot.X+Gy.cndot.Y+Dy(X, Y) (2)
Here, G=(Gx, Gy) are correction coefficients concerning the gain, R=(Rx, Ry) are correction coefficients concerning the rotation, and D=(Dx, Dy) are distortions of higher orders other than the gain and the rotation. In the charged-particle-beam exposure device, a current proportional to the corrected deflection coordinates X'=(X', Y') is applied to the deflector to direct the beam at a desired position X=(X, Y) on the wafer.
When the focusing distance of the lens is changed, the beam cannot be directed to the desired position X. Thus, G, R, and the distortion D(X) must be changed in accordance with the change in the focusing distance.
In order to direct the charged-particle beam at a desired position X on a wafer surface having a given focusing distance (height) f, the correction coefficient G, the correction coefficient R, and the distortion D(X) at various heights f must be measured. In this manner, correction coefficients having the height as a variable, i.e., the correction coefficient G(f), the correction coefficient R(f), and the distortion D(X, f), are obtained. Taking these measurements, however, increases the time for adjusting the beam deflection, and leads to the correction circuit being more complex.
The problem b) will be described below. Instead of using the optical system of FIG. 3 to take the real-time measurement of the height at the time of exposure, reference marks provided on each chip to be exposed can be used for the measurement of the height. Namely, the height of each chip is measured by using the reference marks arranged at four corners of the chip to carry out the focusing and the correction. Since the reference marks have the same predetermined structure irrespective of the chips, the use of such marks allows an easy measurement of the height. In this method, however, the reference marks at the four corners must be detected for the measurement of the height each time the exposure is made. Thus, the processing time is increased. Also, the same as the method of measuring the height in real time, errors in the measurements lead to deviation of the focusing. Further, in case that the heights of the reference marks are not measured for some reason, the focusing on the chip surface cannot be carried out.
In addition to the problems of the mark-position-detection errors and the focusing described above, there is another problem concerning the accuracy of exposed patterns in the charged-particle-beam exposure device.
FIG. 4 is an illustrative drawing for explaining a process of the charged-particle exposure on a wafer. The wafer is divided into areas of a 20-mm square. Here, an IC chip pattern or the like exposed on the wafer generally has a size ranging from a 5-mm square to a 20-mm square. When the IC chips are small, four to nine chips are included together in one area. When the IC chips are large, one chip is included in one area. At the time of exposure, the corrections of the gain, the rotation, the distortion are carried out for each area on the wafer. In general, the exposure data is set for each area unit.
The charged-particle-beam exposure device generally has a main deflector capable of deflecting the beam within a large region and a sub-deflector capable of deflecting the beam at high speed within a small region. The main deflector first directs the beam at a predetermined desired position, and, then, the sub-deflector draws a pattern around the predetermined desired position. In FIG. 4, one area is divided into cells (hereinafter called cell fields), in each of which the main deflector can deflect the beam. When a center point of one cell field is aligned with the axis of the beam optical system, the main deflector can deflect the beam over this cell field. Each cell field has a size of a 1-to-2 mm square. Thus, one area is comprised of about 100 cell fields. Further, the cell field is divided into sub-fields having a size of about a 100-.mu.m square. The sub-deflector can deflect the beam within the sub-field.
The measurements of the deflector-correction data prior to the exposure are conducted within the cell field. Based on the correction data, the correction coefficients of the main-deflector coordinates, the sub-deflector coordinates, and the distortion relative to the stage coordinates are determined. Normally, the correction coefficients of the main deflector are set for each area, and the correction coefficients of the sub-deflector are set for each cell.
The exposure is conducted for a frame, which is a region comprised of a plurality of cell fields arranged in a line. Since the width of the frame is the same as that of a cell field, the beam can be deflected by the main deflector within the width of the frame. For the exposure along a longitudinal direction of the frame, the wafer is successively shifted in the longitudinal direction of the frame through the stage movement. Namely, for the exposure in a transverse direction of the frame, the main deflector deflects the beam for positioning thereof, and the sub-deflector is used for the exposure. For the exposure in the longitudinal direction of the frame, the wafer is successively moved by the stage. After a completion of a one frame exposure, the stage takes a U turn to move the wafer in an opposite direction.
In general, the accuracy of the exposed pattern must be within a 10% tolerance of the exposed pattern. For example, when a 0.15-.mu.m pattern is exposed, the accuracy must be higher than 0.015 .mu.m. In order to achieve this accuracy, the beam correction described above must be precisely conducted. Moreover, there is an effect of a thermal drift of the deflectors during the wafer exposure. Thus, the correction coefficients of the main and sub-deflectors obtained prior to the exposure must be updated during the exposure.
In order to achieve a high precision, therefore, the updating of the correction coefficients must be conducted for each cell or more frequently. The correction coefficients stored in a correction operation circuit must be updated during a break of the exposure such as between the cells or between the sub-fields. Unfortunately, the updating of the correction coefficients is a time consuming process. Thus, frequent updating leads to an increase in exposure time, thereby degrading the performance.
In order to obviate this problem, all the correction coefficients may be calculated and transferred to the correction operation circuit during a time period from a data collection prior to the exposure to the beginning of the exposure. Assuming that 40 coefficients are required for one cell, for example, these 40 coefficients must be calculated for about 4000 points when a 6-inch wafer is used. This means that 1-to-5 seconds are required for the calculation. The data transfer also needs a similar time period. Furthermore, a large-volume memory for storing the correction coefficients is needed in the correction operation circuit. Also, when the correction coefficients are updated during the exposure, all the correction coefficients need to be rewritten after the collection of the correction data.
As described above, there are problems of the mark-position-detection errors, the focusing of the beam, and the setting of the correction coefficients in the related-art charged-particle-beam exposure device. A combination of these problems leads to defects of the generated exposure pattern. When counter measures are taken to avoid these defects, the time required for the adjustment and the exposure is increased, and the device becomes undesirably complex.
Accordingly, there is a need for a device and a method of exposing the charged-particle beam which can create accurate exposure patterns with high productivity.
Also, there is a need for a device and a method of exposing the charged-particle beam which can achieve high-accuracy beam focusing and high-accuracy beam positioning without requiring a long time for beam-deflection adjustment.
Also, there is a need for a device and a method of exposing the charged-particle beam which can accurately detect a mark position.
Also, there is a need for a device and a method of exposing the charged-particle beam which can use correction coefficients provided for small units of areas without sacrificing the exposure processing time.